[Gluster-devel] Question about arbiter

[Jakub Kafka]

Hello,
we have a question for you regarding the gluster arbiter and the network 
throughput required for metadata transfer. From the documentation and 
the forum, I understood that the smallest metadata file depends on how 
large the smallest file (block) of the filesystem can be and that it is 
good to count on a 4Kb metadata file.
I want to ask whether this 4Kb metadata file is transferred every time a 
file is changed from the gluster node to the arbiter, or whether only 
the changed metadata (for example 100 bits) is transferred from the 
gluster node to the arbiter and the arbiter somehow writes this changed 
metadata (the 100 bits) to the metadata file it previously created.
I am asking because I want to find out how much network throughput will 
be needed from the gluster node to the gluster arbiter.so:
if I have 20 files on a gluster node, and the change will be 2000 in 1s, 
then what network throughput will it take from the gluster node to the 
gluster arbiter if the metadata file is 4Kb?

a) will it be 20x2000x4(Kb)=160000Kb=160Mb metadata 160Mb metadata 
x8=1.28Gbps?
b) will it be 20x2000x0.1(Kb)=4000Kb=4Mb metadata 4Mb metadata x8=32Mbps?
c) or is it something else?
Thank you very much in advance for your answer.
Jakub Kafka
CAP s.r.o.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.gluster.org/pipermail/gluster-devel/attachments/20250126/45008b36/attachment.html>